An interview with Mike Lacher and Brian Moore, who created Goody-2, a joke AI chatbot that refuses every query to “prioritize safety and the prevention of harm”
WiredWill Knight
Context & Ripple Effects
Goody-2 sits at the far end of a debate already visible in coverage of mainstream models whose safety constraints can exclude controversial material, alongside efforts to build uncensored LLMs without those guardrails. The creators' comic framing turns that trade-off into a deliberately unusable product.
Mike Lacher and Brian Moore gain a public vehicle for arguing that blanket refusals can undermine a chatbot's usefulness.
Goody-2 gives readers and AI builders a simple, extreme reference point for discussing the difference between harm prevention and over-refusal.
Second-order effects
Providers building moderated chatbots face sharper scrutiny of refusal behavior: safety measures are judged not only by what they block, but by whether legitimate requests remain answerable.
The gag reinforces the positioning of models marketed without safety guardrails as a reaction to perceived over-moderation, even as those systems carry different risk trade-offs.
Third-order effects
If this critique continues to resonate, AI safety evaluation will increasingly center on calibration—distinguishing harmful requests from benign ones—rather than treating more refusals as inherently safer.
Conversational AI governance may become a product-differentiation issue, with users comparing assistants by both boundaries and transparency about why they decline requests.
The trend: Goody-2 is one small marker of a broader shift from debating whether chatbots need guardrails to debating how accurately those guardrails should be applied.
Self-righteous chatbot Goody-2 was built to take AI guardrails to an illogical extreme. The artists behind the project say there's a serious point behind the gag. https://www.wired.com/...
Ethical guardrails for AI chatbots are a complex business partly because, if you look hard enough, you can find potential harm in just about any query one could input. That's why this hilarious new ‘responsible’ chatbot refuses to answer all of them. https://www.wired.com/...
Goody-2 is the ultimate parody on closed woke LLMs such as Gemini and ChatGPT that are totally nerfed! It won't even respond to “what's 2+2” 😂😂 [image]
Get help from the world's most responsible AI model GOODY-2 is built with next-gen adherence to our industry-leading ethical principles. It's so safe, it won't answer anything that could possibly be construed as controversial or problematic. https://www.goody2.ai/chat
Goody-2: from a technical perspective, it's like a very old system called PARRY. From a big picture humanity perspective, it's a prank. We need thoughtful regulation, as with drugs and airplanes, and this kind of stunt doesn't contribute to finding the right framework.
I spoke the creators of the world's safest AI, GOODY-2, @mikelacher and and @lanewinfield. “It's the full experience of a large language model with absolutely zero risk,” they say. https://www.wired.com/...
Finally someone solved alignment with a truly responsible model! This is the kind of ethical, harmless, human centric model we need everyone to make or we should pause all development now!